Add flush element

[bluebear94]

See #4134.

[MDLC01]

I think this should be mentioned on the documentation for place as well.

[tingerrr]

Hey there, this looks great, I figured I'd give it my two cents:

I think this fits better as a sub-element of place as this primarily interacts with place(float: true) elements and nothing else. Similar to figure.caption, it makes no sense to define caption by itself.

I also think this should be included in the default heading show rule, here's my thinking:

• there are almost no documents where a floating figure should be allowed to float past its own heading, this is a sensible default
• we can provide set heading(flush: false) or something similar as an escape hatch, we have precedent for "off-topic" fields like these:
  • heading.outlined being only useful for outline, not heading on its own
  • figure.placement being the inclusion of said floating mechanism for convenience, not necessity, one could wrap figure in place manually
• contrary to the point above, we could not provide this and simply tell the user to recreate the show rule, if they really don't want this
  • the downside of this approach is that adding the flush element to the show rule is trivial, while removing it is not

[bluebear94]

I think this fits better as a sub-element of place as this primarily interacts with place(float: true) elements and nothing else. Similar to figure.caption, it makes no sense to define caption by itself.

I disagree. figure.caption is namespaced within figure because it’s present within a figure element. flush, on the other hand, is not (usually) used within a place element.

I also think this should be included in the default heading show rule

I agree with this, though.

[tingerrr]

I disagree. figure.caption is namespaced within figure because it’s present within a figure element. flush, on the other hand, is not (usually) used within a place element.

I don't think this distinction matters all that much in this case, the semantic relationship is still there, a flush element on its own is useless, it's only useful when used with place(float: true) elements, a subset of place elements at large. assert.eq is also scoped within assert, but used "outside" of it, as it's not an element of course, but it is semantically related to assert (as well as other examples of function scopes).

[PgBiel]

I took a quick look, and at first I think this should work with figure.placement too (please add tests for that), which would, in theory, oppose the argument that this would exclusively affect place. However, I'm also a bit torn on whether this (relatively) small feature should warrant another name in the global namespace. Maybe it's not too bad since #4038 got merged, but worth considering.

Personally I'd be ok with something like #place.flush(), but I'd like to see other suggestions as well.

[tingerrr]

I took a quick look, and at first I think this should work with figure.placement too (please add tests for that), which would, in theory, oppose the argument that this would exclusively affect place.

Well, the implementation of figure simply wraps itself in a place if it has a non-none placement field.

[bluebear94]

Another thing to note is that ‘flush’ has another plausible meaning in the context of typesetting (‘not sticking out’), so perhaps putting it in place’s namespace might be preferable.

[laurmaedje]

My first instinct was also that it should maybe be moved into the place namespace.

[laurmaedje]

I also think this should be included in the default heading show rule

I'm not so sure about this. In a two-column paper, you might have lots of short sections and figures should float past them, I think.

Do we know what LaTeX does by default?

cc: LaTeX and Paper guru @Enivex

[Enivex]

I also think this should be included in the default heading show rule

I'm not so sure about this. In a two-column paper, you might have lots of short sections and figures should float past them, I think.

Do we know what LaTeX does by default?

cc: LaTeX and Paper guru @Enivex

I don't think LaTeX does anything special here. Flushing figures for every heading seems problematic to me.

[PgBiel]

I think we shouldn't force flushing on all headings by default (should be opt-in through show rules).

I additionally don't remember this being the case in LaTeX when I last used it (...which was a long while ago, so take that with a grain of salt :smile:)

[tingerrr]

Yeah, that probably makes sense, in the end it also makes opting out easier.

[laurmaedje]

Thanks!